What is the long term solution to this? In my opinion we should put more emphasis on the data itself, instead of the result. So you do an experiment set to pub

What is the long term solution to this? In my opinion we should put more emphasis on the data itself, instead of the result. So you do an experiment set to publish data, regardless of outcome: the conclusion comes as a sideline, or even in a separate publication altogether (and the exact p-value made just a note). The outcome should be a rebalancing in the bias.